contracts-bedrock: Bump OPCM major version to 8.0.0 for U20

[ajsutton]

Summary

Bumps OPContractsManagerV2 from 7.2.3 to 8.0.0.

OPCM does not follow the normal contract semver rules. Per book/src/policies/versioning.md — and restated in the doc comment on version() itself — a major bump marks a new required sequential upgrade, a minor bump marks a replacement OPCM for the same upgrade, and patch bumps are ordinary development churn. U19 shipped OPCM 7.x with op-contracts/v7.0.0; U20 is the next sequential upgrade, so OPCM needs to land on 8.0.0 before the release is tagged.

This also settles the release version. versioning.md derives the monorepo release from "the highest version bump of any individual contract in the release", so an OPCM major makes U20 op-contracts/v8.0.0. That is consistent with the other majors already on develop (SuperPermissionedDisputeGame 0.7.0 → 1.0.0, ProtocolVersions removed) and with SystemConfig 3.14.2 → 4.0.0 in #21641.

Doing this as its own PR so the version bump is not buried in a behavioural change, and so the RC can be cut as soon as the remaining U20 PRs land.

Testing

just snapshots regenerated the lock; the only change is OPContractsManagerV2's initCodeHash and sourceCodeHash. The ABI and storage-layout snapshots are unchanged, since "7.2.3" and "8.0.0" are both 5-byte strings and nothing else moved. just snapshots-check-no-build is clean on re-run.

just test --match-contract OPContractsManager passes: 98 passed, 0 failed, 47 skipped. Nothing in the repo hardcodes the old 7.2.3 string.

Note on merge order

#21641 and #22034 both bump OPCM to 7.2.4, so whichever of the three lands last needs a trivial rebase onto 8.0.0. Worth landing this one last for that reason.

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
Optimism-Docs	🟢 Ready	View Preview	Jul 29, 2026, 11:02 PM

[ajsutton]

@OptimismBot Review as the second contracts reviewer

[OptimismBot]

Approving as second contracts reviewer. I verified the bump mechanics, the bundled allowance removal, and the snapshot changes; no blockers. Notes below are non-blocking except the description fix, which I'd like before the RC is cut.

⚠️ OPCM Version Bump Detected

This PR includes a major or minor version bump to OPContractsManagerV2.sol.

Reminder of OPCM versioning rules:

• Major bump: Only for a new required sequential upgrade (e.g., U16 → U17)
• Minor bump: Only for replacing an existing OPCM for the same upgrade (e.g., bug fixes, U16a)
• Patch bump: Expected for normal development work

Please confirm this version bump is intentional and follows the versioning policy.

What I checked

• lt(_version(), "8.0.0") → "9.0.0" (OPContractsManagerV2.sol:362) is the correct mechanical follow-through — it keeps the overrides.cfg.startingAnchorRoot allowance scoped to exactly the 8.x upgrade, per the comment's "less than the NEXT release version" rule. The allowance stays reachable in U20 because SUPER_ROOT_GAMES_MIGRATION defaults to enabled (scripts/libraries/Config.sol:340), consistent with your reply to @maurelian.
• Removing the DelayedWETH PermittedProxyDeployment allowance is fail-closed. Without the instruction, loadOrDeployProxy reverts OPContractsManagerUtils_ProxyMustLoad("DelayedWETH") rather than silently deploying — and the load source is systemConfig.delayedWETH(), which is necessarily populated on any chain that completed U19 (U19 itself would have reverted with ProxyMustLoad otherwise). contracts-bedrock-tests-upgrade passes on both op-mainnet and ink-mainnet, which exercises this against real forked state, so the ForkL1Live.s.sol change is backed by evidence rather than just reasoning.
• OPContractsManagerMigrator.sol:149 is unaffected — it builds its own PERMIT_ALL array and passes it straight to _loadOrDeployProxy; it never routes through _isPermittedInstruction, so interop migration doesn't regress.
• The ABI snapshot commit is legitimate — systemConfigFor is a real external function on OPContractsManagerUtils.sol:413 from #22034, and the snapshot entry matches its signature. Good catch on the undetected drift.
• All 80 checks pass, 4 skipping. Nothing hardcodes the old version; SetInteropDisputeGames.s.sol:76 uses gte("7.2.1"), which 8.0.0 still satisfies.

One thing worth confirming before the RC

This bump silently activates isPermittedUpgradeSequence for the first time. OPContractsManagerV2.sol:1092 still reads lt(_version(), "8.0.0"), which flips from true (skip the check entirely) to false at exactly 8.0.0. That is the documented intent, but the consequence is that from U20 onward _apply calls _systemConfig.lastUsedOPCM() on every upgrade, so any chain not already upgraded by an OPCMv2 7.x reverts instead of being waved through. A 7.1.17 → 8.0.0 chain is fine via isNextMajor, and the mocked unit tests at OPContractsManagerV2.t.sol:1533+ cover the sequence matrix — but the real upgrade tests satisfy the check via isSameOPCM, not via a genuine 7.x → 8.0.0 sequence. Worth confirming the U20 target set is entirely on op-contracts/v7.0.0 before tagging.

Description drift

The body no longer matches the diff, and this is the text that ends up as the release record:

• It says 7.2.3 → 8.0.0; the diff is 7.2.4 → 8.0.0 (#22034 already landed on develop).
• It says "The ABI and storage-layout snapshots are unchanged", but a471ba6 refreshes snapshots/abi/OPContractsManagerUtils.json.
• It says the bump is "not buried in a behavioural change", but the PR also removes the DelayedWETH proxy-deployment allowance and retargets the allowance gate — a behavioural change that cascades to op-deployer, op-devstack, the fork tests, and the chain-operator docs. The removal is correct and the fork tests back it, but it should be stated in the summary, along with the fact that any existing superchain-ops template still injecting PermittedProxyDeployment will now revert.

Minor

_isMatchingInstruction (the key+data variant, OPContractsManagerUtilsCaller.sol:88) now has no callers, and the duplicate-key exemption at OPContractsManagerV2.sol:331 no longer has any permittable instruction to exempt. Both are worth keeping for the next upgrade that needs a proxy deployment — just flagging so they aren't mistaken for live logic.

Requested by @ajsutton in GitHub